osFoundry Introduces Team-Like @Mention for Claude Code
osFoundry now allows users to @mention Claude Code or Codex in message channels like a teammate, running coding agents locally to analyze context and solve issues.
Stock photo for illustration only, not from the actual event
- osFoundry integrates Claude Code and Codex support, allowing users to call coding agents via @mention in channels.
- Coding agents run locally on the user's desktop application using a disposable git worktree.
- Users can configure session persistence, choosing between fresh starts or remembering ongoing conversations.
- Built-in safety features restrict prompt injection risks and manage inter-agent communications effectively.
The osFoundry platform has rolled out a new feature that makes interacting with artificial intelligence coding assistants feel just like talking to a human teammate. By typing a mention command inside a message channel, users can trigger tools like Claude Code or Codex to begin working directly on their local machine within a designated repository workspace.
The system reads recent channel messages as context before processing the request and posting the response back under the agent's specific handle. Users can also enable persistence settings so that subsequent mentions retain the ongoing conversational history, allowing the agent to understand previous context without requiring repeated explanations.
Stock photo for illustration only, not from the actual event
Because these coding agents operate directly on local hardware, the feature is exclusively available through the osFoundry desktop application. To set up an agent, users navigate to the channel details sidebar, designate the working folder repository, and add the desired harness from the management panel.
Integrating coding agents directly into team messaging environments bridges the traditional gap between communication tools and local development environments. By allowing developers to trigger repository-level actions straight from a chat thread, teams can streamline collaborative debugging and code reviews without constantly context-switching between separate applications.
The setup panel provides flexible configuration options for session handling, letting users decide whether each interaction should start fresh or maintain a continuous run history. To maintain security, channel messages are fed to the agents as untrusted data with strict length boundaries, preventing any malicious chat inputs from overriding the agent's core instructions.
Teams can also reuse existing agent handles across multiple channels to share conversation history, or create distinct handles to keep separate workspaces isolated. Furthermore, safety controls regulate communications between multiple agents to prevent uncontrolled loops, while providing an easily accessible emergency stop switch for all participants in project rooms.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment